Supplementary Figure 1: Study Design

Night 1 Night 2 Night 3 Night 4

2200 h

Undisturbed Undisturbed Undisturbed Undisturbed

sleep sleep sleep sleep

300 h

tACS tACS tACS tACS

2 Hz/25 Hz sham/12 Hz 6 Hz/40 Hz 70 Hz/100 Hz
3 − 7 REM
during during during during
phases/night
REM sleep REM sleep REM sleep REM sleep
≈ 900 h

Double-blind repeated measures tACS stimulation (30 s duration, each) in REM sleep over 4 non-consecutive
nights. Frequency of stimulation (sham, 2, 6, 12, 25, 40, 70, and 100 Hz) was counterbalanced across subjects
and across nights. The experimenter operating the tACS device did not interact with the subjects. The
experimenter conducting the interviews stayed outside the monitoring room during stimulation, unable to
identify the stimulation condition (for sham stimulation, the push button on the tACS device was activated
but current was not applied). Extensive exploration gave no indication of subjective discomfort or awareness
of stimulation in any subject.

1
Nature Neuroscience: doi:10.1038/nn.3719
 Supplementary Figure 2: Montage and Stimulation Potentials

a b c
Scalp potential Dura potential

st
im
Fp1 Fpz Fp2

ul
~
us
a b
F7 F3 F4 F8
Fz

T3 T4

C3 Cz C4
c TP9 TP10 d
+100 +100
Pz
T5 P3 P4 T6

O1 O2
-100 -100

a) Montage of 22-channel EEG (bandpass filtered: 0.3 – 120 Hz, sampling rate: 512 Hz) and sites of tACS
electrode placement (a, b, c, d). tACS electrodes a and b, as well as c and d were connected pair-wise. The
AC current flowed mainly sagitally between a and c, respectively b and d. The combined impedances of the
paired tACS electrodes were kept below 5 kΩ. EEG electrodes were referenced to Cz instead of mastoids
because of the close proximity of the latter to the tACS electrodes c and d.
b) Measured scalp surface potentials applied during one arbitrarily chosen maximum of a 40 Hz sinusoidal
tACS stimulation in a single subject, with all polarities reverting every 12.5 milliseconds (voltage scale given
in relative units). See Online Methods for details.
c) Mathematically derived dura potential (CSD estimate). The dura potential indicates the effect of the
stimulation current entering the skull, spreading from fronto-temporal to parieto-occipital regions, as well.

2
Nature Neuroscience: doi:10.1038/nn.3719
 Supplementary Figure 3: Empirical Model of Dream Consciousness

Positions on the indicated primary-to-secondary

consciousness axis are based on the logarithm of
ratios of mean scores in lucid and non-lucid dreams.

Consciousness
All factors have been identified as components of
- Insight

Secondary
dream consciousness. Lucid dreams, which are
thought to add elements of secondary consciousness,
are characterized by increased ratings in reflective enhanced
INSIGHT, CONTROL, and DISSOCIATION, and, - Plot Control in lucid
dreams
to a lesser extent, by access to MEMORY, as - Dissociation
well as NEGATIVE and POSITIVE EMOTIONS.
THOUGHT and REALISM do not differentiate
between lucid and non-lucid dreams. The graph is
based on the laboratory scores shown in Fig. 4 of Voss
et al.9 (LuCiD scale). For validation purposes, mean
scores of the dream reports under sham condition in - Negative Emotion
the current sample were compared to those of the - Positive Emotion

Consciousness
original study on the basis of which these factors were - Memory
constructed9 . Scores of both studies compared well,
suggesting reliable subjective ratings in the current Primary - Thought
study (INSIGHT: t = 0.91, p = 0.366; CONTROL: - Sense of Realism
t = 0.46, p = 0.647; THOUGHT: t = 1.20, p = 0.231;
REALISM: t = 1.08, p = 0.283; MEMORY: t = 1.59,
p = 0.113; DISSOCIATION: t = 0.73, p = 0.470;
NEGATIVE EMOTION: t = 1.49, p = 0.139; POS-
ITIVE EMOTION: t = 0.39, p = 0.700; df = 106,
all n.s.).

3
Nature Neuroscience: doi:10.1038/nn.3719
 Supplementary Figure 4: Sample EEG Recordings During tACS

.
Sham stimulation
Awakening

EMG [µV]
100
Filtered
0

−100
Fpz EEG [µV]
EMG
200 Unfiltered Sham
0

{
−200 EEG I II IV
120 140 160 180 200 220 240
Time [s]
12 Hz stimulation
Awakening
EMG [µV]

100
Filtered
0

−100 EMG
Fpz EEG [µV]

200 Unfiltered 12 Hz
on
0
{

{
−200 EEG I II IV
100 120 140 160 180 200 220 240
Time [s]
40 Hz stimulation
Awakening
EMG [µV]

100
Filtered
0

−100 EMG
Fpz EEG [µV]

200 Unfiltered 40 Hz
on
0
{

−200 EEG I II IV
100 150 200 250 300
Time [s]

EEG and EMG data recorded during sham condition (top) and with tACS stimulation of 12 Hz (center), re-
spectively 40 Hz (bottom). The EEG at site Fpz is shown unfiltered while the EMG is filtered to demonstrate
that subject remained in REM sleep throughout stimulation; awakening (sham: t=220 s, 12 Hz: t=195 s,
40 Hz: t=215 s) is signaled by a marked change in the EMG. Note also that the EEG samples shown are
not corrected for ocular artefacts. The horizontal accolades schematically indicate the phases defined in the
main text, namely, before stimulation (I), during stimulation (II), as well as after forced awakening (IV).
During phase II, the tACS current is applied for 30 seconds generating the very large amplitudes indicated
as a solid blue block in the unfiltered EEG (center and bottom frames). Phases I, II and, III (not indicated
here) cover REM sleep, and phase IV corresponds to wakefulness.

4
Nature Neuroscience: doi:10.1038/nn.3719
 Supplementary Figure 5: Grand Average EEG Power vs. Frequency

a phase I b phase I
10 phase II 10 phase II
phase IV phase IV
FFT power [%]

FFT power [%]

1 1

10-1 10-1

10-2 10-2

10-3
Sham 10-3
2 Hz tACS

0 20 40 60 80 0 20 40 60 80
Frequency [Hz] Frequency [Hz]

c phase I d phase I
10 phase II 10 phase II
phase IV phase IV
FFT power [%]

FFT power [%]

1 1

10-1 10-1

10-2 10-2

10-3
6 Hz tACS 10-3
12 Hz tACS

0 20 40 60 80 0 20 40 60 80
Frequency [Hz] Frequency [Hz]

e phase I f phase I
10 phase II 10 phase II
phase IV phase IV
FFT power [%]

FFT power [%]

1 1

10-1 10-1

10-2 10-2

10-3
25 Hz tACS 10-3
40 Hz tACS

0 20 40 60 80 0 20 40 60 80
Frequency [Hz] Frequency [Hz]

phase I 100 Hz tACS phase I

70 Hz tACS
10 phase II 10 phase II
phase IV phase IV
FFT power [%]

FFT power [%]

1 1

10-1 10-1

10-2 10-2

10-3 g 10-3 h
0 30 60 90 120 0 30 60 90 120
Frequency [Hz] Frequency [Hz]

Grand average FFT power was computed as function of the EEG frequency (resolution = 1 Hz) for indicated
stimulation conditions: sham (N=30), 2 Hz (N=31), 6 Hz (N=19), 12 Hz (N=18), 25 Hz (N=26), 40 Hz
(N=44), 70 Hz (N=21), and 100 Hz (N=18). Averaging took place over frontal and temporal electrode sites
(Fp1, Fp2, Fpz, F3, F4, Fz, F7, F8, T3, T4, T5, and T6), over stimulation sequences, and over subjects.
The dips in power mark those frequencies for which a notch filter was applied. Phases I & II correspond to
REM sleep, phase IV to wakefulness.

5
Nature Neuroscience: doi:10.1038/nn.3719
 Supplementary Figure 6: Additional Mean LuCiD Scores vs. Stimulus

Thought Realism
2 4
* **
1 2

0 sham 2 Hz 6 Hz 12 Hz25 Hz 40 Hz 70 Hz100 Hz

0 sham 2 Hz 6 Hz 12 Hz 25 Hz 40 Hz 70 Hz100 Hz
Mean scores

Memory
1
* **
0.5

0 sham 2 Hz 6 Hz 12 Hz25 Hz 40 Hz 70 Hz100 Hz

Negative Emotion Positive Emotion

1.5 2

1
1
0.5

0 sham 2 Hz 6 Hz 12 Hz25 Hz 40 Hz 70 Hz100 Hz

0 sham 2 Hz 6 Hz 12 Hz 25 Hz 40 Hz 70 Hz100 Hz

Stimulus

Mean scores (±1 s.e.) are shown for the LuCiD factors THOUGHT, REALISM, MEMORY, NEGATIVE
EMOTION, and POSITIVE EMOTION. Significant contrasts from MANOVA (N=207) exist only for RE-
ALISM (sham vs. 2 Hz: p = 0.0322, 2 Hz vs. 100 Hz: p = 0.0097) and for MEMORY (sham vs. 70 Hz: p
= 0.0345, 40 Hz vs. 70 Hz: p = 0.0089). In accordance with previous laboratory data9 , THOUGHT and
REALISM are reported in a similar pattern across all stimulation conditions. Albeit not reaching statisti-
cal significance, ratings for both NEGATIVE and POSITIVE EMOTION appear to decrease linearly with
increasing stimulus frequency (as indicated by dashed lines).

6
Nature Neuroscience: doi:10.1038/nn.3719
 Supplementary Table 1: Number of Stimulations and Subsequent Dream Recollections

Nb. of REM sleep Nb. of dreams

Nb. of Awakenings
Stimulus dream reports rated as lucid
tACS during tACS
N fraction [%] N fraction [%]
sham 34 2 30 93.8 0 0
2 Hz 50 10 31 77.5 0 0
6 Hz 46 3 19 44.2 0 0
12 Hz 32 3 18 62.1 1 5.6
25 Hz 42 2 26 65.0 15 57.6
40 Hz 50 2 44 91.7 34 77.3
70 Hz 35 2 21 63.6 3 14.3
100 Hz 35 4 18 58.1 2 11.1

The fraction of dream reports is given with respect to the number of stimulations not leading to spontaneous
awakenings, the fraction of lucid dreams is given with respect to the number of dream reports.

7
Nature Neuroscience: doi:10.1038/nn.3719
 Supplementary Table 2: MANOVA Results on Effects of Stimulation

Wilks’
Factor
Lambda
F df p η2
Stimulation type: 3.29 56, 1039.3 <0.0001 0.12
INSIGHT 4.97 7, 199 <0.0001 0.15
CONTROL 4.68 7, 199 <0.0001 0.14
THOUGHT 0.99 7, 199 0.4335 0.03
REALISM 3.24 7, 199 0.0028 0.10
MEMORY 3.12 7, 199 0.0038 0.10
DISSOCIATION 10.62 7, 199 <0.0001 0.27
NEGATIVE EMOTION 1.72 7, 199 0.1054 0.06
POSITIVE EMOTION 1.35 7, 199 0.2289 0.05

Least-significant difference Bonferroni test for post hoc comparisons:

INSIGHT: CONTROL: DISSOCIATION:

Contrast p Contrast p Contrast p
40 Hz vs. sham 0.0009 25 Hz vs. sham 0.0866 40 Hz vs. sham <0.0001
40 Hz vs. 2 Hz 0.0159 25 Hz vs. 2 Hz 0.0004 40 Hz vs. 2 Hz <0.0001
40 Hz vs. 6 Hz 0.3338 25 Hz vs. 6 Hz 0.0150 40 Hz vs. 6 Hz <0.0001
40 Hz vs. 12 Hz 0.0466 25 Hz vs. 12 Hz 0.0074 40 Hz vs. 12 Hz <0.0001
40 Hz vs. 25 Hz 1.0000 25 Hz vs. 40 Hz 0.0010 40 Hz vs. 25 Hz 1.0000
40 Hz vs. 70 Hz 0.0099 25 Hz vs. 70 Hz 0.0003 40 Hz vs. 70 Hz 0.0016
40 Hz vs. 100 Hz 0.1197 25 Hz vs. 100 Hz 0.0007 40 Hz vs. 100 Hz <0.0001

Multivariate Analysis of Variance results (N=207) on effects of stimulation (sham, 2, 6, 12, 25, 40, 70, and
100 Hz) on subjective ratings of dream consciousness (i.e. the LuCiD scale factors).
INSIGHT: knowing the dream is only a dream while sleeping. CONTROL: being able to change the dream
plot at will. THOUGHT: believing to think logically in the dream. REALISM: the degree to which the
dream feels real. MEMORY: having access to waking memory. DISSOCIATION: 3rd -person perspective.
NEGATIVE EMOTION: feelings of anxiety, anger, or grief. POSITIVE EMOTION: feelings of euphoria or
joy. Also, for validation purposes, mean scores of the dream reports under sham condition in the current
sample were compared to those of the original study9 on the basis of which these factors were constructed
(see also Supplementary Figure 3).

8
Nature Neuroscience: doi:10.1038/nn.3719
 Supplementary Table 3: Spearman Correlation Coefficients

FFT LuCiD Scale Factor

Power Ratio INSIGHT CONTROL THOUGHT REALISM MEMORY DISSOC. NEG. EMO. POS. EMO.
r 0.30*** 0.09 0.12 -0.10* 0.01 0.43*** -0.02 -0.01
40 Hz II/I
p 0.0001 0.2423 0.0789 0.0492 0.8537 2 × 10−8 0.8271 0.8728
r 0.16** 0.09 0.05 -0.03 0.01 0.18** -0.08 0.04
25 Hz II/I
p 0.0098 0.2573 0.5326 0.6571 0.7402 0.0081 0.3908 0.7052

Spearman correlation coefficients r between LuCiD questionnaire data and EEG power ratios (phases II/I)
at fronto-temporal sites in the 25 Hz (22 – 28 Hz) and 40 Hz (37 – 43 Hz) frequency bands, based on 80 s
segments prior to (phase I) and 20 s segments during (phase II) tACS stimulation. N=207, ***: p≤ 0.001,
**: p≤0.01, *: p≤0.05. Power ratios in the 25 Hz band and in the 40 Hz band are correlated with r = 0.22
(p = 0.0012).

9
Nature Neuroscience: doi:10.1038/nn.3719
 Supplementary Table 4: Statistics on Sleep Variables

Night 1 2 3 4 ANOVA
Mean (s.e.) Mean (s.e.) Mean (s.e.) Mean (s.e.) F (df=3, 115) p
TIB [min] 486.39 (12.43) 531.64 (18.41) 509.71 (9.00) 521.50 (17.03) 1.93 0.121
SPT [min] 467.11 (13.16) 514.79 (17.31) 491.93 (8.98) 499.14 (16.04) 2.10 0.098
TST [min] 330.70 (12.13) 375.52 (12.68) 343.57 (9.01) 357.71 (13.11) 2.67 0.053
Sleep eff. [%] 67.99 (1.37) 70.63 (1.32) 67.40 (1.06) 68.59 (1.11) 1.57 0.226
SOL [min] 19.27 (1.27) 16.86 (1.68) 17.79 (0.99) 22.36 (1.82) 2.56 0.090
% WASO 23.84 (1.30) 21.78 (1.40) 26.86 (1.59) 22.60 (1.18) 2.56 0.090
% light sleep 44.49 (1.12) 44.85 (0.65) 42.67 (1.03) 41.10 (1.48) 2.35 0.078
% SWS 18.51 (0.78) 19.04 (1.17) 18.48 (1.32) 22.26 (1.25) 2.54 0.106
% REM sleep 7.32 (0.49) 9.14 (0.68) 8.50 (0.91) 8.46 (0.36) 1.57 0.238

TIB (in min): Time in bed, TST (in min): total sleep time, SPT (in min): sleep period time, sleep efficiency:
TST/TIB x 100, SOL: sleep onset time (stage 1 sleep), WASO: wake after sleep onset, SWS: slow wave sleep.
Sleep variables do not differ significantly across nights (ANOVA, all p >0.05).

10
Nature Neuroscience: doi:10.1038/nn.3719